Men who hate women: The misogyny of involuntarily celibate men

Incels and networked misogyny

Incels see themselves as victims of lookism, a form of prejudice and discrimination that orients to physical attractiveness (Dion et al., 1972; Halpin, 2022). Incel discourse is characterized by the “black pill,” the fatalistic acceptance that women’s mating preferences are structured by lookism, and, because of their unattractiveness, incels will be “forever alone” (Glace et al., 2021; Preston et al., 2021).

Incels are part of a larger manosphere, a group of male-oriented websites that share the belief that men face systemic discrimination (Bates, 2021; Ging, 2019; Marwick and Caplan, 2018; Van Valkenburgh, 2021). Incels and similar communities create affordances (Boyd, 2010) for the villainization of feminism, cyberhate, and networked misogyny (Halpin, 2022; Halpin and Richard, 2021; Jaki et al., 2019; Jane, 2018; Jones et al., 2020; Marwick and Caplan, 2018; Moloney and Love, 2019; O’Donnell and Shor, 2022; O’Malley et al., 2020; Pelzer et al., 2021; Preston et al., 2021). Building on such research, we use computational methods to demonstrate the ubiquity of misogyny on incels.is.

Incels are not a monolith. Incels are diverse in relation to social class, age, and race/ethnicity (Halpin and Richard, 2021). While the site we examine bans women, some women identify as incels and some researchers further suggest distinguishing between incels and misogynistic incels (e.g. Arbeit et al., 2022; Kelly et al., 2021). In this article, rather than focusing on misogynistic or non-misogynistic incels, we examine misogynistic discourse produced by users of incels.is. We refer to this group specifically when we use the term “incel” throughout our analyses and discussion.

Gender, sexism, and intersectionality

Our analyses are informed by social scientific theories on intersectionality, masculinity, and sexism. First, we draw on intersectionality (Collins, 1990), which is a framework demonstrating how multiple axes of oppression overlap. From this perspective, misogyny is never just misogyny, but also shaped by other forms of oppression, such as racism. In this article, we ask: how is incel misogyny both racist and racialized? We address this question by comparing misogynistic terms that target White women (e.g. “Stacy”) with terms that target Women of Color (e.g. “big black vagina”). As we argue, incel misogyny orients to a matrix of domination (Collins, 1990), wherein all women are denigrated, but women of color are doubly denigrated through a combination of racism and sexism.

Second, we draw on Connell’s (1995) conceptualization of hegemonic and subordinate masculinity. Hegemonic masculinity is not a trait or a type of man, but rather is a set of practices that legitimate and maintain unequal gender relations, including unequal relations between groups of men (Connell, 1995; Messerschmidt, 2018). While there are many dimensions to this concept (Connell, 1995; Messerschmidt, 2018), we discuss hegemonic masculinity as practices that advance the subjugation of women.

In contrast to hegemonic masculinity, subordinate masculinity refers to practices that have been “exiled” from masculinity. Connell (1995) argues that men who participate in subordinate masculinity have their masculine legitimacy questioned (e.g. gay men). Connell likewise asserts that men who do not participate in heterosexual sex can be positioned, or position themselves, as subordinate. In this article, we discuss subordinate masculinity in relation to incels’ assertions that they are failed men because of lookism and their lack of participation in heterosexual sex (see also Halpin, 2022).

Research has tied incels to both hegemonic and subordinate masculinity. Some of these arguments (e.g. Ging, 2019; Glace et al., 2021) assert that incels practice hybrid masculinity, which is a blending of different masculine practices (e.g. hegemonic and subordinate masculinity; see Bridges and Pascoe, 2014). For instance, Halpin (2022) argues that incels weaponize their subordination, as they strategically position themselves as subordinate, “failed men” to justify their hegemonic practices, such as endorsing interpersonal and systemic violence against women. Situating misogyny as a hegemonic practice that perpetuates unequal gender relations (Connell, 1995; Messerschmidt, 2018), we ask: to what extent do incels participate in hegemonic masculinity? We address this question by documenting the usage of misogynistic terms both in incel.is threads and users’ post histories, as well as examining if users become more misogynistic in relation to posting frequency. We argue that incels enact hybrid masculinity, as they self-identify as subordinate men while ubiquitously participating in misogyny.

Third, we use Connell’s (1987) concept of “emphasized femininity,” a counterpart to hegemonic masculinity, that is defined by nurturing, empathizing, and taking a submissive orientation toward men and their needs. The relationship between emphasized femininity and hegemonic masculinity is asymmetrical, as hegemonic masculinity dominates the gender order. Contrastive to emphasized femininity, Schippers (2007) discusses pariah femininities, which are socially undesirable forms of femininity. Building on these tensions between culturally valued and shunned femininities, we ask two questions: (1) how do incels talk about different groups of women? and (2) how do incels use misogynistic terms in comparison with neutral terms for women (e.g., “women”)? We address these questions by comparing the usage of different types of misogynistic terms (e.g. “Stacy,” “noodlewhore”) and compare the use of misogynistic terms and non-misogynistic terms. We argue that incels rank order women, but nonetheless treat all femininities as pariah femininities.

Finally, while numerous studies detail misogyny in the manosphere, these studies do not leverage the theory of ambivalent sexism (Glick and Fiske, 1996). Glick and Fiske (1996) suggest that sexism is a “special case of prejudice,” as sexist heterosexual men pursue intimacy with women, while simultaneously feeling antipathy toward them. The theory identifies two types of sexism: (1) hostile sexism, which is aggressive misogyny that situates women as controlling, manipulative, and subhuman, and (2) benevolent sexism, which endorses complementary gender roles and frames women as requiring men’s protection. Researchers argue that hostile and benevolent sexism are highly correlated, fitting “hand in glove,” with one another (Glick and Fiske, 1996). However, recent studies suggest that both incels (Halpin and Richard, 2021) and men who see themselves as unattractive (e.g. Bosson et al., 2022) endorse hostile sexism, but have decreased endorsement for benevolent sexism. In this article, we ask: how does incels’ participation in hostile sexism compare to their participation in benevolent sexism? We address this question by using “girl” as an indicator for benevolent sexism, and comparing its frequency to the use of hostile sexist terminology that dehumanizes women (discussed below). Given incels prolific hostile sexism, we argue that they evidence how the “hand” of hostile sexism slips out of the “glove” of benevolent sexism.

Procedure

We collected all posts (e.g. text posted by a user) that appeared on the incels.is discussion board (“Inceldom Discussion”) from 8 November 2017 until 16 April 2021 (N = 3,686,110). We employed custom scraping scripts ¹ to extract all public post data. This script automatically traversed and downloaded the corresponding HTML for every page of posts within each thread of “Inceldom Discussion.” Individual comments and user information were then extracted from the HTML and saved as thread-specific text files. The extracted data include post text along with the associated participant user IDs, posting time, thread titles, threads, and the respective order of posts in threads. Our analyses do not include posts that were deleted before we completed our data collection and excludes text quoting other posts in the same thread.

Analytic approach

We use computational methods to collect and describe incel misogyny, while using social science theories to inform our research questions, as well as interpret and discuss our results. In this sense, we aim to leverage computational analyses, while addressing critiques that computational approaches are overly descriptive and/or atheoretical (Edelmann et al., 2020; Nelson, 2020). Likewise, drawing on previous work (Nelson, 2020), we employ an inductive approach that involves first creating a glossary of misogynistic terms, then developing and employing regular expressions ² to capture all references to these terms, and finally abductively analyzing (Halpin and Richard, 2021) our data in relation to social scientific theories.

Most analyses of the incel community focus on discrete comments or threads. In addition to analyzing comments and threads, we analyze misogyny in the first post, and the post title. We display data at these three levels to describe how misogynistic discussions can develop over the course of a thread (e.g. do discussions begin or become misogynistic?). In further contrast to previous analyses, our dataset also allows us to characterize misogynistic language in relation to a user’s posting frequency (see Table 1 and Figure 2).

Since incels use considerable jargon (e.g. terms like “foid”), we first developed a glossary of misogynistic terms to cross-reference against discussion board posts. Our glossary includes the misogynistic terms listed in the incels.is glossary (Incels.wiki, 2021), as well as the misogynistic glossary on Hatebase.org (2021), as the incels.is glossary omits some common misogynistic terms (e.g. “bitch,” “cunt”). In addition to using these glossaries, we also included misogynistic terms that we documented in our prior qualitative analyses (Halpin, 2022; Halpin and Richard, 2021; Preston et al., 2021) that are both used by incels and missing from their glossary (e.g. “currywhore”).

Drawing on the social science theories reviewed above, we assigned a label to each misogynistic term to differentiate between types of misogyny (e.g. “Racist Misogyny”). We employed a decision rule to prevent double-counting terms by prioritizing their category assignment in the following order: “Racist Misogyny,” “Stacy/Becky,” “Girl,” “Female,” “Foid,” and “Other Misogyny.” In practical terms, our decision rules mean that a term like “noodlewhore” is assigned to “Racist Misogyny” and the word “whore” that follows “noodle” is not double-counted into the “Other Misogyny” category. ³

Our category “Racist Misogyny” captures misogynistic terms that are also racist (e.g. calling Asian women “noodlewhores”), while “Stacy” or “Becky,” are misogynistic terms for White women (see also Menzie, 2022). We coded “female” and “girl” separately because these terms can be used non-misogynistically or misogynistically. “Female” can operate as a misogynistic microaggression (e.g. Sue et al., 2007), while we treat “girl” as an indicator of benevolent sexism (Glick and Fiske, 1996). We discuss the coding of “female” and “girl” in the “Fidelity checks” section below. We created a separate category for variations of “foid” (e.g. “foid,” “femoid”), as they are, by far, the most prolifically used misogynistic terms among incels. Any misogynistic term that is not captured by these categories (e.g. “bitch,” “roastie”) is placed in the category “Other Misogyny.”

For comparative purposes, we also identified “Neutral Terms” for women (e.g. “woman”). While we code these words as “neutral” because they are non-misogynistic ways to refer to women, this analysis does not detail the broader context of word use. Consequently, our analysis likely underestimates the extent of incel misogyny (see “Fidelity checks” section).

While our coding rules orient to social science theories (e.g. Collins, 1990; Connell, 1995), our rules are also a limitation, and different rules would produce different analyses and answer different questions. For instance, we could have created a category of incel-specific misogynistic terms (e.g. “foid,” “noodlewhoore”) to potentially address how incels do or do not adopt in-group terminology in relation to their posting history. However, because so many incel-specific terms are also racialized (e.g. “Stacy,” “currywhore”), our prohibition on double-coding terms means we would have to had sacrificed our analyses of racist misogyny, which was a focus of this project. Nonetheless, decision rules for coding must be made to avoid double-counting terms, and different coding approaches are an opportunity for future research.

Fidelity checks

Given the size of our dataset, our analysis contains both false positives (i.e. non-misogynistic terms identified as misogynistic) and false negatives (i.e. misogynistic terms not identified as misogynistic). To improve the accuracy of our categorizations, we manually reviewed a random sample of 100 instances of each term (e.g. “foid”) and compared the results produced by our script with an alternative script. ⁴ This validity check identified several terms with false positives. For example, we repaired queries that falsely coded “hole” as a misogynistic term when used in “asshole” or “black hole.”

We also considered that some participants might employ misogynistic terms to critique other participants (e.g. “you shouldn’t call women whores”). To investigate the prevalence of such uses, we reviewed a random sample of 500 posts that were coded as “Other Misogyny.” All 500 posts were misogynistic, and none contained instances of discouraging misogyny, suggesting such a practice is either exceedingly rare or absent.

Finally, “woman,” “girl,” and “female” can be used in misogynistic and non-misogynistic ways. We reviewed a random sample of 1000 comments using “woman/women,” finding that 528 (52.8%) were misogynistic (e.g. “when a woman is raped, this takes her power away”). We also reviewed a random sample of 500 comments that include “girl” and another 500 for “female.” We found that 362 (72.4%) of the uses of “girl” were misogynistic. Here, our coding is informed by the concept of benevolent sexism and instances of referring to adult women as “girls” were counted as misogyny. We also found that 463 (92.6%) uses of “female” were misogynistic, wherein users referred to women with the noun “female” (e.g. “females avoid me”).

Misogyny in discussion board threads

Misogynistic terms are ubiquitously used in incels.is threads. Summarizing across all six of the misogynistic subcategories that we document in Table 1 (i.e. “Total Misogyny”), discussion board participants used misogynistic terms 960,487 times during the 42-month period of study. Nearly one fifth (17.3%) of all comments contain at least one misogynistic term. While 17.7% of thread titles contain misogynistic terms, nearly double that number (39.4%) of the first posts contain a misogynistic term, meaning that misogynistic first posts might have non-misogynistic titles. When looking at the threads in their entirety, 82.3% of threads on incels.is contain at least one misogynistic term. Put differently, only 17.7% of threads do not contain one of the misogynistic terms in our glossary.

In contrast, we record 402,045 uses of “Neutral Terms” (e.g. “woman”). We find that 7.3% of posts, 7.9% of threads titles, and 20.8% of thread first posts, and 57.8% of all threads contain “Neutral Terms.” These findings demonstrate that misogynistic terms are used 2.4 times more frequently than non-misogynistic terms, meaning that incels refer to women with misogynistic terms considerably more frequently than they use non-misogynistic terms. Moreover, as 82.3% of threads contain misogynistic terms, whereas 57.8% of threads contain neutral terms for women, a substantial portion of threads only refer to women with misogynistic language. These findings demonstrate that threads on incels.is are characterized by misogyny (see Supplemental Appendix 2 for the 25 most frequently used misogynistic terms).

Misogyny in participants’ post histories

The vast majority of participants on incels.is use misogynistic terms. We find that 81.2% of participants used at least one misogynistic term during the study period (see Table 1). In contrast, 67.7% of participants used neutral terms for women (e.g. “women”), meaning that some users exclusively refer to women using misogynistic terms. Participants on incels.is are on average 3.3 times more likely to use a misogynistic term than a non-misogynistic term for women. In terms of participants’ frequency of using misogynistic terms, 45.6% of participants made five or fewer posts containing a misogynistic term (see Table 2). However, participants who used misogynistic terms less frequently also had lower overall posting rates, with users who did not use any misogynistic terms creating an average of 3.8 posts. This suggests that non-misogynistic incels (e.g. Arbeit et al., 2022) rarely contribute to discussions on incels.is. Likewise, we also identify some truly prolific misogynists, documenting 67 users who have produced more than 2000 posts that contain at least one misogynistic term. Most participants are between these two poles, with 54.4% of participants making at least six posts that contain at least one misogynistic term.

We also find that users with a higher number of total posts were more likely to have made at least one comment containing a misogynistic term. Figure 1 shows that 53.5% of participants with 1–10 posts used at least one misogynistic term, whereas 92.2% of participants with 11–25 posts have used at least one misogynistic term. This trend continues, such that nearly 100% of participants with 51 posts or more have used at least one misogynistic term. While Figure 1 reports cross-sectional data, Figure 2 reports longitudinal data to assess the relationship between use of misogynistic terms and participation on incels.is.

To evaluate the relationship reported in Figure 2, we fit a simple linear regression model predicting the per-participant change in average monthly term usage with the number of months a participant has spent on the forum as the predictor for each of the categories we report in Table 1 (e.g. “Racist Misogyny”). To be able to analyze a longitudinal trajectory, we excluded all users with posting activity in fewer than 3 months to ensure that our results characterize users who participated on incel.is over time. For the remaining participants (N = 3043), average monthly term use was calculated as the number of times a participant used terms belonging to each category divided by their monthly post count. To obtain the change in their average use over time, the difference between a participant’s average usage in their first and last active month was calculated. An ordinary least squares (OLS) model was then fitted for each term category (e.g. “Racist Misogyny”) without an intercept using the statsmodel v0.13.2 library.

As can be seen in Figure 2, these models suggest that there was no clear positive or negative overall relationship between a participants’ use of the terms in any of the categories and the duration of their participation on incels.is. While we observed a significant but weak negative association for “Total Misogyny” (p = .0064, β = −.0019, R² = .33%), “Stacy/Becky” (p = .0131, β = −.003, R² = .26%), “Girl” (p < .000, β = −.0014, R² = .64%), and “Other Misogyny” (p = .029, β = −.0007, R² = .20%) these models explained <1% of the variance (R²) relative to a constant model with no association.

Together, Figures 1 and 2 suggest that overall participation in misogyny might be tied to a survivorship bias among incels.is users, wherein users who do not produce misogynistic content are less likely to maintain participation, whereas users who maintain membership arrive, rather than become, misogynistic.

Our analyses also allow us to compare misogyny in threads with misogyny in participants’ post histories. For instance, “Racist Misogyny” appears in 8.7% of threads; however, one third (30%) of participants have used a “Racist Misogyny” term at least once. Put simply, by comparing thread data with user post history data, we see that users might not frequently start racist misogynistic threads, but users pile on and join a racist misogynistic bandwagon when such content appears. This finding likewise suggests that while thread-based analyses are a popular approach, such analyses might understate forms of incel misogyny.

Types of misogyny

While misogyny is ubiquitous on incels.is, incel misogyny is not uniform. Drawing on social science theories (e.g. Collins, 1990), our findings also describe different types of incel misogyny. Participants used “Racist Misogyny” terms (e.g. “noodlewhore”) more than 33,000 times. “Racist Misogyny” terms appear in 8.7% of threads, while 30% of participants have such terms in their post histories. The incel terms “Stacy” and “Becky” are also racialized, as they refer to attractive White women who incels situate as rejecting them and selecting attractive men (Menzie, 2022). “Stacy” and “Becky” are not compliments, as “Stacy” is depicted as a “vain . . . entitled whore,” while Becky is used to describe as exploitative women with low self-esteem (Incels.wiki, 2022a, 2022b). Incels used “Stacy/Becky” 41,000 times, with these terms appearing in 13.6% of threads and 38.5% of participants’ post histories. These findings suggest that a substantial portion of incel misogyny is racist or racialized, which we discuss below in relation to intersectionality and conceptualizations of femininity.

Incels’ use “female” just over 115,000 times, with the term appearing in 31.7% of threads and 48.5% of participants’ post histories. Participants employ “girl” more than 200,000 times, with this term appearing in 41.9% of threads and 63.2% of participants’ post histories. Participants use variations of the misogynistic incel term “foid” more than 250,000 times, with this term appearing in 50.1% of threads and 50% of users’ post histories.

Our category “Other Misogyny” consists of misogynistic terms that do not fit into any of the preceding categories (e.g. “Racist Misogyny”). We record more than 315,000 instances of this type of misogyny, which includes common terms like “cunt” and “whore,” as well as terms included in the incels.is glossary, such as “war pig” and “landwhale.” Terms in the “Other Misogyny” category appear in 6.5% of posts, 56.6% of threads, and 65% of users’ post histories.

We also examined incels’ participation in specific types of misogyny in relation to their posting frequency. As Figure 3 shows, users with a higher number of posts are more likely to participate in these types of misogyny. For example, use of “Racist Misogyny” and “Stacy/Becky” increases from almost 0% among the users with 1–10 posts to over 80% for participants with 501–600 posts. While this cross-sectional data might be taken to suggest that users become socialized in relation to racist misogyny as they post more on incels.is, Figure 2 does not support this interpretation, demonstrating that there is no or weak statistically significant relationships between forum activity and any type of misogyny. Once again, these data suggest that the men who arrive to incels.is engaging in racism and misogyny are also more likely to become long-term contributors to the discussion board, rather than have their post content change over time.